Bakar and Tahir (2009) in their paper used multiple
linear regression technique and simulated neural network
techniques for predicting bank performance. ROA was
used as dependent variable of bank performance and
seven variables including liquidity, credit risk, cost to
income ratio, size and concentration ratio, were used as
independent variables.
They concluded that neural network method
outperforms the multiple linear regression method
however it need clarification on the factor used and they
noted that multiple linear regressions, not withstanding its
limitations, can be used as a simple tool to study the
linear relationship between the dependent variable and
independent variables.
Neceur (2003) using a sample of ten Tunisian banks
from 1980 to 2000 and a panel linear regression model,
reported a strong positive impact of capitalization to ROA.
There are number of studies, which examine the bank
performance using CAMEL framework, which is the latest
model of financial analysis.